Determinant-based Pose Estimation Solution to Handle Both Point and Line Outliers

This article aims to address the problem of accurately estimating camera pose using both 2-D/3-D point and line features, also known as perspective- n -point/line. In practical applications, point and line outlier correspondences can be introduced together by image algorithm mismatches and pose estimation accuracy may decrease. Therefore, a unified feature framework is proposed in this article to remove both types of feature outliers. Specifically, the nonlinear projections of the two different features are first transformed into a unified linear system. Then, two determinant-based error functions are defined based on the linear system, wherein the point and line functions are solved to obtain the features derivation distributions by the quadratic equation and the Rayleigh quotient, respectively. Furthermore, the feature outliers rate is decreased by analyzing the derivation distributions, and the null space method is applied for removing all the outliers. Finally, the proposed algorithm effectiveness is validated through simulations using the synthetic and open datasets.